Co-folding models hold immense potential for allosteric drug discovery, but have been severely hampered by their systematic bias toward orthosteric ligand binding. While fragment screening has been proposed for allosteric binding site discovery, we show that co-folding models still suffer from memorization in which chemically simpler fragments also default to canonical orthosteric binding sites. To overcome these limitations, we introduce CAFE (Co-folding Approach for Fragment Exploration), a co-folding protocol that uses competitive orthosteric blockers to divert fragments into non-canonical sites as illustrated here with the Boltz-2 co-folding model. Using ADP as an orthosteric blocker for the kinase family, we find CAFE substantially increases the allosteric binding site exploration for fragments, with notably strong absolute binding free energies that match or exceed those of known crystallographic poses, without post-hoc refinement of the Boltz-2 prediction. We also show that CAFE identifies cryptic binding pockets undetected by conventional pocket prediction tools, some of which are more thermodynamically favorable than the allosteric or orthosteric pockets. To demonstrate generality, we apply CAFE using Type I orthosteric blockers for kinase proteins, known orthosteric ligands as blockers for non-kinase proteins in the RAS-MAPK signaling pathway, and for virtual screening campaigns using fragment libraries for new fragments that selectively engage allosteric and cryptic binding sites. CAFE establishes orthosteric blocking and fragment screening as a training-free, inference-time protocol that helps overcome some of the limitations of current co-folding models while elevating their great promise for allosteric and cryptic binding drug discovery.
Cryptic pockets are transient ligand-binding sites that remain hidden in apo protein structures and become accessible only through conformational change, making them difficult to identify and exploit in structure-based ligand discovery. Although recent AI/ML approaches can identify residues associated with cryptic pocket formation, they do not directly yield the corresponding holo-like conformations. Enhanced-sampling and mixed-solvent strategies can promote pocket opening, but often under non-native conditions and without demonstrating stable ligand-bound holo-like states. Here, we present a computational framework in which residues with high cryptic-pocket propensity are used as mutation handles to perturb the free-energy landscape and expose ligand-compatible open states even in the absence of ligand binding. We establish this strategy on TEM-1 β-lactamase, a canonical cryptic-pocket system, and show that mutation-induced perturbations can shift the conformational ensemble toward an open pocket state prior to ligand binding. The resulting open state supports ligand binding in both docked structures and unbiased simulations, and the ligand remains stably bound after reversion to the wild-type sequence, consistent with recovery of a holo-like wild-type conformation. Importantly, the residues targeted for mutation are not the principal determinants of the ligand interactions observed in the bound state, indicating that their role is to reshape the conformational landscape and promote access to an open, ligand-compatible pocket rather than to form the binding interface itself. We further generalize this framework to LfrR, FtsZ, and Bombyx mori pheromone-binding protein, where mutation of predicted cryptic residues likewise generates open conformations capable of supporting ligand binding. Together, these results show that cryptic residue predictions can be used not only to identify hidden binding sites but also to recover holo-like conformations from apo structures, providing a practical framework for studying and targeting cryptic pockets.
Reshob Routh, Mithun Radhakrishna· Journal of Chemical Informat...· 0 citations
Allostery offers a powerful route to regulate protein function and expands drug discovery beyond the orthosteric paradigm. By acting at sites distinct from the active site, allosteric modulators can achieve greater selectivity, reduce off-target effects, and overcome resistance. The discovery of the cryptic switch-II pocket of KRAS, which turned a long-"undruggable" oncoprotein into a clinically validated target, exemplifies this promise. Yet allosteric drug discovery is demanding: it requires not only identifying a suitable, often transient pocket, but also demonstrating that this pocket is functionally coupled to the active site, and then translating that mechanistic insight into design. This perspective surveys the computational strategies addressing each of these challenges in turn: sequence, structure, and machine-learning-based methods for locating allosteric and cryptic sites; network and dynamical analyses for mapping communication pathways; and enhanced-sampling and generative deep-learning approaches for rational modulator design. Throughout, we emphasise a central theme: that generative AI delivers speed and breadth, while physics-based simulation supplies thermodynamic rigour, and that their integration, rather than either alone, defines the most promising path forward. Together with experimental validation, these advances are rapidly expanding our ability to exploit allosteric regulation in therapeutics.
Sutanu Mukhopadhyay, Suman Chakrabarty· Chemical Communications· 0 citations
This study provides potential lead compounds for the design of small-molecule allosteric drugs targeting class B1 GPCRs and performs conformational sampling and combined dynamic pocket detection algorithms, MDpocket and FTMove, to identify six characteristic cryptic pockets within the dynamic trajectories.
Zhi Dong, Long Cheng, Qingxin Shi et al.· International Journal of Bio...· 0 citations
Boltz2ESI is introduced, an end-to-end framework that predicts enzyme–substrate interactions by leveraging structural knowledge learned by a biomolecular foundation model and consistently outperforms state-of-the-art sequence-based and rigid-docking approaches.
Xiwei Cheng, Seonghwan Seo, C. Huh et al.· bioRxiv· 0 citations
Allosteric modulation of Abl kinase via its myristoyl binding pocket is a promising strategy for drug discovery, either aiming at inhibition to treat Chronic Myeloid Leukemia, or activation, under investigation for breast cancer therapy. Although activators and inhibitors can be differentiated by the induced α-helix-I conformation, in silico classification has been proven challenging. Our study distinguishes these classes effectively by integrating multiple computational approaches that account for the conformational plasticity of the regulatory α-helix-I. By evaluating traditional molecular docking, co-folding, and molecular dynamics simulations with methodology benchmarking, we use these methods to uncover mechanistic principles of allosteric modulation. Docking highlighted potentially stabilizing C–F interactions with deep-pocket residues, while co-folding correctly predicted helix bending for inhibitors and outperformed docking in virtual screening. Molecular dynamics revealed that the α-helix-I samples multiple possible conformations and uncovered motions consistent with dynamic coupling between helix bending and long-range restraint of the activation loop, a nuanced mechanism that static models could not elucidate. This study provides a validated framework that combines efficient classification with mechanistic analysis by combining molecular docking, co-folding, and molecular dynamics. Our approach aids in identifying myristoyl pocket ligands, differentiating inhibitors from activators, and suggesting allosteric principles that govern their function, paving the way for more rational design of next-generation, function-specific Abl modulators.
P. T. T. F. Leite, P. O. Fernandes, D. M. Martins et al.· Journal of Chemical Informat...· 0 citations
Protein kinases are critical drug targets, requiring therapeutics that can modulate their active and inactive conformational states. While cofolding models can generate global folds directly from kinase sequences and ligand SMILES strings, these models have not yet been tested on their ability to recover ligand-induced-fit conformational states of the kinase proteins. Here, we introduce KinConfBench, a curated benchmark of 2225 high-quality human kinase chains to evaluate the ability of four state-of-the-art cofolding models—Boltz-2, Chai-1, Protenix, and RoseTTAFold-All-Atom—to recover both canonical and rare conformational states. We show that geometric success metrics of a ligand pose in the active site do not correlate strongly with the correct kinase conformational state, motivating a new set of dynamical benchmarks for assessing cofolding models. While all four cofolding models achieve ~60–80% prediction accuracy for kinase conformational classification, they exhibit severe mode collapse when performing multiple inferences, show negligible structural diversity in sampling induced-fit motions, and display a prevalent “apo-drift” in which most cofolding models predominantly predict the kinase to be in its ligand-free state. Our results highlight that capturing ligand-induced protein conformational diversity, not just geometric fit, is critical for next-generation structure-based drug discovery.
Kunyang Sun, T. Head-Gordon· npj Drug Discovery· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
